Reading a GEM with a VLSI pixel ASIC used as a direct charge collecting anode Ronaldo Bellazzini INFN Pisa Vienna February 16-21 2004
The GEM amplifier The most interesting feature of the Gas Electron Multiplier (GEM) is the possibility of a full decoupling of the charge amplification structure from the charge collection and read-out structure.
Two-dimensional Readout Concepts Amplifying structure and read-out structure can be optimized independently of each other. The electron charge is collected on strips, pixels or pads on the read-out board. A fast signal can be detected on the top GEM electrode for triggering or energy discrimination. Cartesian Small angle Fast trigger Y-coordinate Pixels X-coordinate
Pixel read-out: Advantages By organizing the read-out plane in a multi-pixel pattern it will be possible to get a true 2D (imaging) capability. A high granularity of the read-out plane would also allow to preserve the intrinsic resolving power of the device and its high rate capability that otherwise would be unavoidably lost by using a conventional projective read-out approach. For example: GEM rate capability ~ 10 6 /mm 2 s A strip 20 cm long at 400 µm pitch should withstand a global rate of 100 MHz, prohibitive for standard, low noise electronics. Ambiguity problems in projective read-out when dealing with multi-tracks or multihits events.
Pixel read-out: an example, the PCB approach X-RAY Polarimetry Angle and amount of polarization is computed from the angular distribution of the photoelectron tracks, reconstructed by a finely segmented gas detector. Read out plane GEM pitch: 90 µm GEM holes diameters: 45 µm, 60 µm Read out pitch: 260 µm Absorption gap thickness: 6 mm 512 electronic channels from a few mm 2 active area are individually read out by means of a multi-layer PCB fan out
Polarimetric sensitivity 5.9 KeV unpolarized source 5.4 KeV polarized source C 2 ( φ ) = A + B cos ( φ φ ) 0 Modulation factor = (Cmax Cmin)/ (Cmax + Cmin) 50% at 6 KeV MDP ( n ) 1 MDP scales as: ( µ ε ) n = εµ S σ σ 2 εs + B AT ( µε ) 1 1 for bright sources for faint sources
The limits of the PCB approach The fan-out which connects the segmented anode (collecting the charge) to the front end electronics is the real bottleneck! Technological constraints limit the maximum number of independent electronics channels (~ 1000 @ ~ 200 µm pitch). Crosstalk between adjacent channels (signals traveling close to each other for several cm). Not negligible noise (high input capacitance to the preamplifiers).
A further technological step: the CMOS VLSI approach If the pixel size is small (below 100µm) and the number of pixels is large (above 1000) it is virtually impossible to bring the signal charge from the individual pixel to a chain of external read-out electronics even by using the advanced, fine-line, multi-layer, PCB technology. When it is not possible to bring out the signal charge to external, peripheral electronics than it is the electronics that has to be brought in to the individual pixel! A CMOS full custom pixel array used directly as the charge collecting anode of the GEM has been designed, produced and it is currently under test. Advantages: asynchronous, fast, low noise, honeycomb array design, no problems in the realization of the fan out to front-end electronics.
Our first prototype of pixel read out has 2101 channels and 80 µm pitch. Each microscopic pixel is fully covered by a hexagonal metal electrode realized using the top layer of a 6 layers, 0.35 µm CMOS technology. This charge collecting pad is individually connected to a full chain of nuclear type electronics (preamplifier, shaper amplifier, sample and hold, multiplexer) which is built immediately below it making use of the remaining 5 active layers.
The collecting anode/read-out chip pixel electronics dimension: 80 µm x 80 µm in an exagonal array, comprehensive of preamplifier/shaper, S/H and routing (serial read-out) for each pixel number of pixels: 2101
Electronics conceptual design ~3.5 microseconds shaping time 100 e - ENC ( very small detector capacitance) dynamic range: 0.2-20 fc power consumption: around 100 µwatt/pixel external trigger (from the GEM) for parallel S/H on all the channels ADC after S/H: external, flash 400 µs read-out time (with 5 MHz system clock)
Pixel Event read-out timing As long as the MaxHold signal is low the shaped pulse of the selected pixel can be observed at the analog output.
Analog out timing characteristics A pixel is selected by introducing a token into the shift register
PIXIE: the PIXel Imager Experiment Detector and associated electronics are the same thing!
Exploded mechanical assembly transfer gap ceramic package entrance window absorbtion gap VLSI chip (collecting anode) GEM
The analog signal 3.5 µs Analog output (60000 electrons) Write signal (no Maxhold) Analog output (1000 electrons) Analog output (6000 electrons) Analog output (60000 electrons) Shaper output S/H analog output Maxhold signal Automatic search of the maximum of the signal within a 10 µsec window after an asyncronous external trigger (from the TOP GEM)
Noise measurement Single channel analog output (few k random triggers) Pedestal Analog out (ADC counts) Noise RMS: 1.8 mv ~ 100 e - ENC (electronics gain is ~ 100 mv/fc): sensitive to the single primary electron with a gas gain ~1000 (easily achievable with a single GEM).
Channel response uniformity and X-talk 3 pixels strobed with 1 V signal (1000 ADC cnts)
Internal calibration system and addressing capability Detector response to 20 mv calibration signal (~1000 electrons) injected in a subset of pixels to create the experiment Logo. Excellent response uniformity even before any attempt of calibration.
Tracks reconstruction 1) The track is recorded by the PIXel Imager 2) Baricenter evaluation 3) Reconstruction of the principal axis of the track: maximization of the second moment of charge distribution 4) Reconstruction of the conversion point: major second moment (track length) + third moment along the principal axis (asymmetry of charge release) 5) Reconstruction of emission direction: pixels are weighted according to the distance from conversion point.
Tracks morphology Auger Electron Bragg Peak Raw data: less than 40000 electrons subdivided on 46 pixels!
Some events
Monte Carlo simulation: propagation of the photoelectron Photoelectron transport code originally developed by D. Joy for electron microscopy (particularly accurate at low energies) - adapted for the transport in gas mixtures. Preferred emission direction Photoelectron and Auger propagation in Ne/DME 80/20
Full detector simulation Generation (photoelectron + Auger) Propagation (SS_MOTT) Creation and diffusion of primary ionization (Maxwell, Garfield, Magboltz) Gas multiplication Digitization Pixel Representation
Data analisys: Ne/DME 80/20 1 Atm
Conclusions A system in which the GEM foil, the absorption gap and the entrance window are assembled directly over a CMOS chip die has been developed. The ASIC itself becomes at the same time, the charge collecting anode and the pixelized read-out of a MicroPattern Gas Detector. For the first time the full electronics chain and the detector are completely integrated. At a gain of 1000 a high sensitivity to single primary electron detection is reached. No problems found up to now in operating the system under HV and in gas environment. An astronomical X-ray Polarimeter application has been presented. Depending on pixel and die size, electronics shaping time, analog vs. digital read-out, counting vs. integrating mode, many others applications can be envisaged. Final design will have 16-32 k channels and 60-70 microns pixel size. This would open new directions in gas detector read-out, bringing the field to the same level of integration of solid state detectors.